Boxplot Tutorial

Looking at Metabolites: Boxplots

1. We need to import the dataset of all metabolites and their frequencies in the soil samples (indicated as each variable/column)

data <- read.csv("EMSL_PNNL.csv", header=TRUE)

2. From a t-test of control vs heated groups of every metabolite, the mineral metabolites and organic metabolites that were significantly different between the two conditions were the following:

3. Let’s look at only one signficant metabolite: glyceric acid

data_ga <- data[36,]

4. Data Wrangling

ggplot is an effective package in R that allows for plotting data in a clean and pretty manner. This includes boxplots, violin plots, line graphs, histograms, and more.

ggplot plots two variables in a dataset as x and y of a grid. Thus, the x and y must be columns in a dataset.

In our current dataset “data_ga”, we want to plot 4 boxplots with approximately 5-10 values in each boxplot.

So, the x column must include character values of “Mineral Control”, “Mineral Heated”, “Organic Control”, and “Organic Heated”. The y column must include the frquencies.

We must gather the wide-looking dataset into the long-looking dataset. The function to do so is gather() (which is a function under the package tidyverse or subpackage tidyr) and more information can be found here:https://www.dropbox.com/sh/d5thhny56crrfir/AABLBA44Ngmlus--rLzlh7zGa/3.%20Data%20Wrangling?dl=0&preview=dataWrangling.pdf&subfolder_nav_tracking=1

referenced from Dr. Douglas Rice of the Political Science Department; Winter 2020 Workshop

library(tidyverse)

## Warning: package 'tidyverse' was built under R version 3.6.2

## -- Attaching packages ------------------------------------------------------------------ tidyverse 1.3.0 --

## v ggplot2 3.2.1     v purrr   0.3.3
## v tibble  2.1.3     v dplyr   0.8.3
## v tidyr   1.0.2     v stringr 1.4.0
## v readr   1.3.1     v forcats 0.4.0

## Warning: package 'tidyr' was built under R version 3.6.2

## Warning: package 'purrr' was built under R version 3.6.2

## -- Conflicts --------------------------------------------------------------------- tidyverse_conflicts() --
## x dplyr::filter() masks stats::filter()
## x dplyr::lag()    masks stats::lag()

data_ga_wrangled <- data_ga %>% gather("M_C_4_AB",  "M_C_7_AB", "M_C_12_AB",    "M_C_14_AB",    "M_C_19_AB",    "M_C_30_AB",    "M_H_2_AB", "M_H_2B",   "M_H_4_AB", "M_H_11_AB",    "M_H_17_AB",    "M_H_26_AB",    "M_H_28_AB",    "M_H_30_AB",    "M_H_32_AB",    "M_H_34_AB",    "O_C_4_AB", "O_C_12_AB",    "O_C_14_AB",    "O_C_19_AB",    "O_C_27_AB",    "O_C_34_AB",    "O_H_2_AB", "O_H_4_AB", "O_H_11_AB",    "O_H_17_AB",    "O_H_32_AB", key="type", value="frequencies")

The ‘values’ in the x column we called “type” are all different names even though all the ones starting with M_C are mineral controlled. Let’s rename the values so that we only end up with 4 different values.

data_ga_wrangled$type <- c("Mineral Controlled", "Mineral Controlled", "Mineral Controlled", "Mineral Controlled", "Mineral Controlled", "Mineral Controlled", "Mineral Heated", "Mineral Heated", "Mineral Heated", "Mineral Heated", "Mineral Heated", "Mineral Heated", "Mineral Heated", "Mineral Heated", "Mineral Heated", "Mineral Heated", "Organic Controlled", "Organic Controlled", "Organic Controlled", "Organic Controlled", "Organic Controlled", "Organic Controlled", "Organic Heated", "Organic Heated", "Organic Heated", "Organic Heated", "Organic Heated")
head(data_ga_wrangled)

##      Metabolite               type frequencies
## 1 glyceric acid Mineral Controlled    129353.9
## 2 glyceric acid Mineral Controlled    143556.3
## 3 glyceric acid Mineral Controlled    213930.1
## 4 glyceric acid Mineral Controlled    271421.3
## 5 glyceric acid Mineral Controlled    337871.4
## 6 glyceric acid Mineral Controlled    258831.6

5. ggplot

the syntax of ggplot is first the data then its aesthetics (aes) which are essentially the variables of the graph. Then additional layers can be added “+”.

Below, the fundamental ggplot is: ggplot(data, aes(x=type, y=frequencies, fill=type)))

The ‘fill’ is the color of each boxplot. The color will be grouped up by the “type” of value (aka the x variable).

Then, the added layers is “geom_boxplot()” which makes the output a boxplot and “geom_point()” which outputs points on the already added boxplots.

ga <- ggplot(data_ga_wrangled, aes(type, frequencies, fill=type))
ga +
  geom_boxplot() + 
  geom_point()

6. ggplot additional details

Next, let’s add color to the boxplots and label (labs):

To fill the boxplots with specific colors manually, hex codes can be used (and can be found online). Note that we’ve already filled the boxplots using the ggplot aesthetic “fill” by type (above) and we are now overlaying the default fills with manual colors.

boxga <- ga + geom_boxplot() + geom_point() +
  scale_fill_manual(values=c("#808080", "#B43757", "#808080", "#830300")) +
  labs(title="glyceric acid Frequencies")
boxga

Additional manipulations can be done to the ggplot including changing the axis labels, font sizes, colors, legend placement, etc.

Here is the ggplot cheatsheat:

https://rstudio.com/wp-content/uploads/2015/03/ggplot2-cheatsheet.pdf

7. Transformations

A great explanation on why transformations are necessary in Biological statistics:

http://www.biostathandbook.com/transformation.html

Common transformations:

Log - allows incredibly large numbers to be transformed into digestable small numbers

Square-root - It is weaker than the log or cubed root and is used to reduce right skewedness

Square - To reduce right skewedness

referenced from:

(https://fmwww.bc.edu/repec/bocode/t/transint.html) and Professor Roy Thompson FRSE of The University of Edinburgh.

Visual for Transformations

from (https://www.statisticssolutions.com/transforming-data-for-normality/)

Utilizing log transformations:

boxga + scale_y_continuous(trans = 'log')

The y-axis ticks are innacruate to the boxes. To change that, the breaks and labels must be corrected using trans_breaks and trans_format. These functions can be found in the package “scales”.

library(scales)

## 
## Attaching package: 'scales'

## The following object is masked from 'package:purrr':
## 
##     discard

## The following object is masked from 'package:readr':
## 
##     col_factor

boxga + scale_y_log10(breaks = trans_breaks("log10", function(x) 10^x),
              labels = trans_format("log10", math_format(.x)))

more information and a tutorial: https://r4ds.had.co.nz/transform.html

8. Violin plots

Similar to boxplots but includes a ‘rotated kernel density’ plot on each side rather than a box.

Plotting the same data as a violin plot:

violin_ga <- ga + geom_violin()
violin_ga

You can display the median and quartiles in the form of a boxplot

median_v_ga <- violin_ga + geom_boxplot(width=0.1, col="red")
median_v_ga

Or the mean and standard deviation

mean_v_ga <- violin_ga + stat_summary(fun.data=mean_sdl, geom="pointrange", color="red")
mean_v_ga

With such small datasets, the standard deviation lines are unsurprising. Personally, of these two options (displaying mean or median), I prefer the violin plot with the median.

More info is here: http://www.sthda.com/english/wiki/ggplot2-violin-plot-quick-start-guide-r-software-and-data-visualization